Effects of Selenium Administration on Blood Lipids: A Systematic Review and Dose–Response Meta-Analysis of Experimental Human Studies

Abstract

Context

Overexposure to the essential trace element selenium has been associated with adverse metabolic and cardiovascular outcomes, hypertension, and diabetes. However, dose–response meta-analyses analyzing the effects of selenium administration on the lipid profile in experimental human studies are lacking.

Objective

Through a restricted cubic spline regression meta-analysis, the dose–response relation between the dose of selenium administered or blood selenium concentrations at the end of the trials and changes over time in blood lipids, ie, total, high-density lipoprotein (HDL) and low-density lipoprotein (LDL) cholesterol, and triglycerides was assessed.

Data Sources

Searches were performed on PubMed, Web of Science, Embase, and the Cochrane Library from inception up to January 11, 2025 to identify randomized controlled trials (RCTs) investigating the impact of selenium supplementation on blood lipid profiles among adults.

Data Extraction

A total of 27 eligible RCTs that enrolled healthy individuals, pregnant individuals, and participants with specific health conditions were identified and the relevant data was extracted.

Data Analysis

Dose–response analysis indicated that selenium administration at and above 200 µg/day decreased HDL and LDL cholesterol and increased triglyceride levels. Blood selenium concentrations at the end of the trial above approximately 150 µg/L were positively associated with triglyceride and LDL cholesterol concentrations, and inversely associated with HDL cholesterol. Inorganic selenium supplementation showed stronger associations than organic selenium. At the lowest levels of baseline intake, selenium supplementation appeared instead to have beneficial effects on the lipid profile, with an overall indication of U-shaped curves, apart from HDL-cholesterol. The adverse effects of selenium were stronger in studies involving healthy participants as compared with unhealthy participants and pregnant females, in those having a longer duration of the intervention, particularly more than 3 months, and in European populations at selenium intake levels of above 300 µg/day.

Conclusions

In this dose–response meta-analysis of experimental human studies, an adverse effect of selenium administration on blood lipids at levels around or above the current upper level of intake was observed.

Systematic Review Registration

PROSPERO registration No. CRD42022380432.

INTRODUCTION

Selenium is a trace element with nutritional and toxicological properties, depending on its dose and chemical species.^1–3 The main source of selenium in humans is diet, largely determined by the amount of this element in the soil.⁴ Seafood and fish, meats, cereals, mushrooms, and Brazilian nuts are foods rich in selenium.^5,6 The recommended dietary intake of selenium ranges from 34 to 70 µg/day, depending on the criteria used for such an assessment.^7,8 The physiological functions of selenium are mediated by selenoproteins such as selenoprotein P (SEPP1) and glutathione peroxidase (GPx), with transport and antioxidant properties.⁹ Conversely, toxic properties are ascribed to selenoproteins themselves, to various inorganic and organic selenium species, and to selenocompounds, at high levels of exposure.^8,10

Some investigators suggested during the 1970s that selenium may contribute to the prevention of cancer, cardiovascular diseases, and other chronic diseases, at both nutritional and supranutritional exposure levels, on the basis of some observational evidence, biological plausibility, and the first randomized controlled trial (RCT) carried out with selenium.^7,11–18 However, large and methodologically stronger RCTs have recently failed to confirm any beneficial effect of selenium for cancer and cardiovascular diseases.^19,20 These experimental human studies also raised concerns about selenium toxicity, including an increased risk of metabolic and high-grade prostate cancer.²⁰ Such effects add to the symptoms and signs of acute and chronic selenium intoxication, such as vomiting, diarrhea, pain, nausea, a garlic-like odor on the breath, nail abnormalities, alopecia, and dermatitis.^4,21 Adverse metabolic effects of chronic low-dose selenium overexposure may contribute to obesity, hypertension, dyslipidemia, non-alcoholic fatty liver disease, and type 2 diabetes in particular, as supported by some experimental and nonexperimental studies among humans.^21–29 However, the extent to which these adverse effects are mediated or accompanied by an adverse effect on lipids is not well defined. Selenium, in fact, is known to regulate progenitor cell proliferation, adipocyte differentiation and maturation, as well as lipolysis.²² Previous animal experiments have reported increased activity of LDL-receptor, the expression of mRNA, and lipid accumulation after selenium supplementation.³⁰ A recent meta-analysis showed that selenium increases levels of low-density lipoprotein (LDL) cholesterol and systolic blood pressure.³¹ However, to our knowledge, no dose–response meta-analysis has been carried out to assess the pattern of association between the entire range of selenium exposure and lipid profile, with particular reference to the possibility of nonlinear relations.

Within the context of active investigation into the effects of selenium on the lipid profile in the human, a meta-analysis aiming to assess the dose–response effect of selenium administration on lipid profiles in RCTs was conducted. To do that, a newly developed statistical methodology was used, ie, the one-stage approach³² based on restricted cubic spline models, which enables the use of results from all studies when at least 2 levels of exposure are available.³³

METHODS

This systematic review was conducted following the PRISMA recommendations.³⁴ The protocol for this study has been registered on PROSPERO (No. CRD42022380432). The GRADE approach with 5 domains (imprecision, inconsistency, indirectness, risk of bias, and publication bias) was used to rate the certainty of the evidence.³⁵

Literature Search

Searches for the relevant literature in PubMed, Web of Science, Embase, and the Cochrane Library, with no language restrictions, from inception up to January 11, 2025 were performed. The search terms “selenium”, “lipid profile”, “LDL”, “very-low-density lipoprotein cholesterol (VLDL)”, “high-density lipoprotein (HDL) cholesterol”, “cholesterol”, “triglycerides”, and “randomized clinical trials” were used, by setting the research question according to the PICOS statement (Population, Intervention, Comparator(s), Outcomes, and Study design) (Table 1). Detailed search strategies are provided in Figure 1. The gray literature was also retrieved, according to the EUnetHTA guidelines.³⁶ In this review, only experimental studies with randomized and controlled designs in adults were considered, encompassing oral or parenteral administration of selenium through dietary supplements, and whenever possible including the evaluation of internal exposure using blood biomarkers, ie, serum or plasma levels.³⁷ The end point of interest was the lipid profile: Cholesterol, HDL, LDL, and triglycerides.

Retrieved articles were imported into the Rayyan QCRI online application and duplicates were removed. Two authors (T.U. and G.F.) independently screened publication titles and abstracts and evaluated full-text publications for inclusion in the review. Full-text articles were included when both reviewers agreed that the inclusion criteria were met. Agreement between the authors who screened the articles was higher than 90% (Kohen’s κ > 0.90). In cases of disagreement, both authors performed a second review of the full text to determine eligibility for inclusion through a consensus-based discussion. If the 2 authors still disagreed, a third author (T.F.) was sought to resolve the disagreement. Table S1 provides a fully detailed list of the full-text articles excluded, with related reasons for exclusion.

Data Extraction

Two authors (T.U. and G.F.), with the help of a third author (T.F.), extracted the following data from each eligible study: (1) first author name; (2) publication year; (3) country; (4) dose and duration of selenium supplements, selenium concentrations before and after the intervention (when these data were available), and the difference between the intervention and control groups at the end of the intervention; (5) type of population; (6) outcome of interest (levels of cholesterol, HDL, LDL, triglycerides) before and after the intervention; (7) number of participants; (8) duration of the study; (9) blood levels of selenium before and after intervention; and (10) sample type (serum or plasma). For 1 study,³⁸ the dose of selenium supplementation was converted to the effective dose of Na-selenite administered per day (⁠$100 µg selenium:333 µg Na-selenite=500 µg selenium:x$⁠). For 5 studies,^38–42 cholesterol and triglyceride levels were converted from mmol/L to mg/dL, using as conversion factors 38.67 and 88.57, respectively.⁴³ Two studies^44,45 provided conversion factors. Mean or median values along with standard deviation (SD), standard error (SE), or interquartile range (IQR) were extracted. For data conversion from SD to IQR, the formula proposed by Cochrane was used, ie, IQR/1.35.⁴⁶ If not already reported, the mean difference of the change in selenium plasma levels was calculated as Mt—Mc, were Mt is the mean in the selenium treatment group and Mc is the mean in the control/placebo group. To impute the SD of the change from baseline for the experimental intervention, the following formula according to Cochrane methodology⁴⁷ was used:

Risk of Bias

Two authors (T.F. and T.U.) independently assessed the risk of bias of the included studies using the specific tool for randomized trials, RoB 2.0 (Table S2).⁴⁸ Disagreement was resolved with the help of a third author (M.V.). Five domains, with each of them rated as having a low, moderate, or high risk of bias, were analyzed, as follows: (1) “bias arising from the randomization process”, assessing both allocation concealment and random allocation; (2) “bias due to deviations from intended interventions”, assessing blinding of participants and personnel, and reporting of selenium levels before and after the intervention; (3) “bias due to missing outcome data”, assessing whether authors did not report the results anticipated in the trial protocol or methods; (4) “bias in the measurement of outcome”, assessing the modality of the lipid profile evaluation; and (5) “bias in selection of reported results”, based on information on the trial registration to avoid selection of the results produced. If 2 domains were considered to have some concerns, the overall judgment resulted in “some concerns.” Otherwise, the studies were considered to be at low risk of bias.

Data Analysis

In our meta-analysis, a first comparison of the blood lipids in the intervention groups, ie, the selenium-supplemented participants, with the blood lipids of participants receiving the placebo (or the lowest amount of selenium, if a placebo was not available). Forest plots were used via a random-effects model in order to consider possible heterogeneity of the results of the included studies.⁴⁷ A dose–response meta-analysis was then performed, by assessing changes in blood lipids during the follow-up, according to the dose of selenium administered or to the circulating selenium concentration achieved at the end of the intervention. The weighted mean difference (MD) and 95% confidence interval (CI) of 4 categories of blood lipids was estimated, due to homogeneity in the units of measurements of the outcomes. Meta-analysis was performed when at least 2 studies were available for each end point. For the dose–response meta-analysis, a restricted cubic spline model based on the 1-stage methodology (an approach for dose–response meta-analysis based on a weighted mixed-effects model that enables use of the results from all studies when at least 2 levels of exposure are available) was used.^49–52 The estimation method was based on the restricted maximum-likelihood random-effects model.^32,53 The restricted cubic spline model was generated using 3 knots at fixed percentiles (10th, 50th, and 90th) of blood selenium concentrations. For each analysis, different reference doses were used. For selenium supplementation, the reference value was set to 0 µg/day, since the control group did not receive any selenium supplementation. For modeling the relationship between circulating selenium concentrations at the end of the trial and lipid end points, a reference point of 105 µg/L of blood selenium was used, corresponding to an intake of approximately 70 µg/day.⁵⁴ Whenever possible, stratified analyses according to population characteristics (ie, sex and health status), country region (Europe vs Asia), type of selenium compounds (“organic” vs “inorganic”), and study duration (≤12 weeks vs >12 weeks) were performed. The dose–response meta-analysis was only computed when 3 or more studies were available. Meta regression analyses using the trial duration as an independent variable were also performed. For all data analyses, the results were reported and interpreted based on an evaluation of the magnitude, direction, and precision of the effect estimates, rather than binary significance testing.^55–57

Egger‘s test to assess potential publication bias via contour-enhanced funnel plot asymmetry were used.⁵⁸ The presence of heterogeneity between the effect sizes was assessed using I² and τ² tests,⁵⁹ and sensitivity analyses were performed assessing study-specific curves.³⁵ Leave-one-out sensitivity analyses were then performed to quantify the impact of potential study outliers on the estimation of the overall effect size.

All analyses were carried out using the “drmeta,” “meta,” “mkspline” routines of Stata software (StataMP 18.0, StataCorp LLC, College Station, TX, 2023).

Certainty of the Evidence

The overall certainty of the evidence was assessed using the GRADE approach.⁶⁰ Taking into account the PICOS question, the certainty was assessed for changes in MDs in lipid profile levels due to selenium supplementation yielded by the dose–response analysis in all studies. The GRADEPro GDT (https://gradepro.org) was used to present the certainty assessment and to summarize the findings in tabular form.

RESULTS

In Figure 1, the flow chart presents the stages of study retrieval and the assessment of eligibility. After the removal of 528 duplicate publications, 654 potentially eligible publications were identified from the literature search, from which 607 publications were further excluded based on title/abstract screening. After full-text evaluation, 47 studies were evaluated for inclusion in the final analysis. Of these, 20 studies were excluded for the following reasons: Wrong supplementation if the study used other drugs/supplements in addition to selenium; wrong outcome, if not evaluating blood lipid profile or not evaluating their levels both before and after treatment; and if not a RCT. The numbers of publications excluded after the full-text evaluations, with reasons for the exclusions, are reported in Table S1.

The main characteristics of the 27 studies included in the meta-analysis are presented in Table 2. Most trials (N = 21) were carried out in Iran,^{40,42,61–79} while the remaining 6 trials were carried out in Europe, namely 2 in Denmark,^39,41 and 1 each in Finland,⁴⁴ Germany,⁸⁰ the United Kingdom,⁴⁵ and the Czech Republic.³⁸ Overall, the trials included 2958 participants (1830 in the selenium treatment groups and 1128 in the control groups). Nine studies recruited only female participants,^{40,61,62,68,71,72,74,78,79} and 1 included only male healthy subjects.⁴¹ The other 16 studies enrolled male and female participants, but none performed sex-stratified analysis. The duration of the trials ranged from 2 to 24 weeks (median = 12 weeks). Selenium supplementation occurred in the inorganic form using sodium selenite in 4 trials, while organic selenium was administered in the form of selenium-enriched yeast in 20 studies. In 3 trials, selenium was administered in both inorganic and organic forms.^41,69

For intervention with the inorganic selenium, the most common supplemented dose was 200 µg/day, ranging from 96 µg/day⁴⁴ to 480 µg/day.⁴¹ For selenized-yeast, the administered dose ranged from 100 to 300 µg/day. One study did not specify the dose administered.⁷³ Blood selenium concentrations (reported as either plasma or serum concentrations) at the beginning and end of the trial were available in 10 studies, showing an increase in selenium concentrations after the intervention in every trial.^{38,39,41,44,45,64,66,73,78,80}

Four studies were conducted on generally healthy populations,^39,41,44,45 and 3 trials among pregnant females,^62,72,78 whereas the majority of trials (N = 19) were carried out in patients with chronic diseases. Three trials recruited participants diagnosed with polycystic ovary syndrome,^40,74,79 5 trials enrolled hemodialysis patients,^{63,64,73,75,76} 5 articles engaged patients affected by cardiovascular disease,^{42,67,69,70,80} 2 studies involved women with central obesity,^61,68 1 study was conducted on patients affected by Systemic Inflammatory Response Syndrome (SIRS)/sepsis,³⁸ 2 studies involved diabetic patients,^65,66 1 study focused on patients with cervical intraepithelial neoplasia,⁷¹ and 1 trial was conducted in patients with Alzheimer’s dementia.⁷⁷ Regarding the end points, the majority of studies reported blood concentrations of total cholesterol (N = 25), HDL cholesterol (N = 24), LDL cholesterol (N = 21), and triglycerides (N = 23). Two trials reported blood concentrations of non-HDL cholesterol.^39,45

Almost all studies were deemed to be at low risk of bias; 5 were considered to have moderate risk of bias,^{38,44,74,78,80} while none were considered at high risk of bias (Table S2). Bias in the domain “Effect of assignment to intervention” was the most common, generally due to a missing measurement/reporting of blood selenium concentrations before and after supplementation, in both the placebo and treatment groups.

In the forest plot, comprising the 27 studies and comparing the highest vs lowest exposure category in each study, selenium supplementation was associated with an overall reduction in total cholesterol (MD: −2.06; 95% CI −4.71, 0.58). A smaller but very imprecise decrease was observed for triglycerides (MD: −0.96; 95% CI −5.96, 4.04). No substantial effect was observed for LDL (MD: −0.52; 95% CI −3.23, 2.18) and HDL cholesterol (MD: 0.26; 95% CI −0.86, 1.37) (Figures S1-S4).

Figure 2 shows the dose–response meta-analysis evaluating the effects of selenium supplementation on lipid profiles. Selenium supplementation above 200 µg/day was associated with increased levels of total cholesterol (N = 25) and triglycerides (N = 23) and decreased levels of LDL (N = 21) and HDL (N = 24) cholesterol. Specifically, for total cholesterol, there was a U-shaped association, with lower total cholesterol levels observed for selenium concentrations between 100 and 200 µg/day, after which their levels slightly and linearly increased. Conversely, selenium supplementation showed an almost linear inverse association with LDL cholesterol above 100 µg/day of intervention. The association with HDL cholesterol was slightly inversely U-shaped, with higher HDL levels at selenium supplementation of approximately 100 µg/day (Figure 2).

When evaluating blood selenium concentrations achieved by the end of the study in the 10 trials reporting this measure, there was a positive association with total (N = 10) and LDL cholesterol (N = 9) and triglyceride (N = 8) levels, although with different patterns of association. For total and LDL cholesterol, the association was U-shaped, with an inflection point of the curve at approximately 125 µg/L. For triglycerides, the association was J-shaped, the inflection point being located at approximately 100 µg/L. HDL cholesterol was slightly linearly inversely associated with selenium levels (N = 9) (Figure 3).

In analyses stratified according to participant characteristics, negative effects and increased MDs of total cholesterol (MD: 3.14; 95% CI −8.56, 14.85), LDL (MD: 2.89; 95% CI −6.96, 12.73), and triglycerides (MD: 9.89; 95% CI −8.70, 28.49) were observed only among pregnant females compared with healthy/unhealthy subjects. Among unhealthy subjects and pregnant females, selenium supplementation was associated with increased HDL cholesterol levels, while detrimental effects were observed in the healthy subject category (MD: −2.13; 95% CI −3.72, −0.54) (Figures S5-S8). With regards to sex, evidence for a detrimental effect of selenium administration was observed on total cholesterol (MD: 0.45; 95% CI −5.97, 6.87) and LDL cholesterol (MD: 1.63; 95% CI −4.01, 7.27) among females, and on HDL cholesterol (MD: −3.11; 95% CI −8.18, 1.96) among males. No sex differences were noted on triglyceride concentrations (Figures S9-S12). When analyses were stratified by type of selenium supplementation (organic vs inorganic), inorganic selenium generally showed the most detrimental effects on total (MD: 21.39; 95% CI −22.00, 64.79) and LDL cholesterol levels (MD: −2.32; 95% CI −2.34, 6.98), though the number of studies based on inorganic forms (N = 4) was much lower than the number of trials administering organic selenium species (N = 20) (Figures S13-S16). When stratifying by geographic region, some differences emerged between studies conducted in Asia and Europe (Figures S17-S20). Opposite and slightly adverse effects were observed for HDL and LDL cholesterol levels in European (MD: −1.80; 95% CI −3.24, −0.36 and MD: 1.55; 95% CI −3.00, 6.11, respectively) compared with Asian populations (MD: 1.42; 95% CI 0.12, 2.72 and MD: −1.37; 95% CI −4.58, 1.85, respectively). Trial duration also seemed to affect the end point, since less favorable effects emerged in particular for LDL (MD: 3.01; 95% CI −10.20, 16.22) and HDL cholesterol (MD: −0.50; 95% CI −2.68, 1.67) and triglyceride concentrations (MD: 0.53; 95% CI −28.85, 29.90) with a longer period of intervention, ie, more than 3 months of selenium supplementation (Figures S21-S24). Further assessment of the role of trial duration using meta-regression analysis confirmed such an association (Figure S25).

When feasible, dose–response meta-analyses stratified by population characteristics (healthy/unhealthy/pregnancy status and sex), type of selenium administration (supplementation with organic or inorganic selenium), and country were also performed. Summary estimates were generally imprecise. Among healthy subjects, selenium supplementation showed a similar effect on total cholesterol compared with the overall population, while the impact on HDL cholesterol levels seemed to be detrimental and inverse. In unhealthy subjects, the association with total and HDL cholesterol were flat compared with the overall population, showing no substantial effects, while it had an inverted U-shape for LDL cholesterol with decreasing concentrations above 200 µg/day, and an inverse relation for triglycerides. The associations appeared almost null for pregnant females (Figure S26).

The highest dosages of selenium supplementation had slightly detrimental effects on total and HDL cholesterol and triglyceride in studies including both males and females, while demonstrating no adverse effects on LDL cholesterol levels. Studies including only females showed inverted U-shaped relations with all examined lipid measures (Figure S27).

Dose–response meta-analysis according to the specific selenium form administered was feasible only for studies using organic supplementation (N = 22). Exclusion of studies using only inorganic supplementation (N = 4) showed a pattern of U-shaped association for total cholesterol and an inverse U-shaped association for triglycerides and LDL cholesterol. Conversely, HDL cholesterol showed a slight increase due to supplementation with organic selenium (Figure S28).

Finally, stratified analysis by region showed a marginal effect of selenium supplementation in Asian populations (N = 20) and detrimental effects at >300 µg/day in European populations (N = 5), when considering total cholesterol and triglycerides as an outcome, although the maximum intake in Asia was lower than in Europe. A slight positive association was observed with HDL cholesterol, while an inverted U-shaped association was observed with LDL cholesterol (Figure S29).

Sensitivity analysis assessing study-specific curves for selenium supplementation showed very high variation for HDL cholesterol (Figures S30-S31). Leave-one-out analysis showed substantial homogeneity of the results, with some exceptions, including a lower MD for LDL cholesterol and triglycerides when excluding 1 group for Schnabel et al (2008)⁸⁰ and Tara et al (2010),⁷⁸ respectively (Figures S32-S35). Similarly, exclusion of the single study not administering selenium through the oral route³⁸ did not affect the overall estimate (Figure S32).

Counter-enhanced funnel plots for publication bias are shown in Figures S36-S39, indicating some evidence of asymmetric distribution for LDL cholesterol, suggesting possible occurrence of some publication bias for this end point.

Finally, GRADE assessment showed generally a low certainty of the evidence (since not all studies were carried out in the general population) and very serious imprecision, although the presence of dose–response increased the level of certainty (Table S3).

DISCUSSION

This dose–response meta-analysis, the first so far carried out to the best of our knowledge, gave an indication that selenium supplementation can adversely affect lipid profiles, particularly HDL cholesterol and triglycerides, and partially LDL cholesterol. These results agree with those of some observational studies, in which circulating selenium concentrations were positively associated with cholesterol (total and LDL) and triglycerides concentrations, and with the risk of dyslipidemia.^28,30,81 The results of this meta-analysis suggested a threshold for a detrimental effect of selenium on total and HDL cholesterol and triglycerides at approximately 200 µg/day, while opposite and inconsistent results emerged for LDL cholesterol. It is noteworthy that, for this end point, trials with the longest duration of intervention indicated an adverse effect of selenium supplementation. In addition, no such inconsistency was noted when blood selenium concentrations were used to assess actual exposure status in the trial participants, since all the investigated end points consistently indicated adverse effects of circulating selenium levels at approximately 100 µg/L and above, corresponding to approximately 70 µg of daily selenium intake.⁵⁴ For HDL cholesterol, adverse effects occurred starting at 50 µg/Se/L. This would imply that the upper level of selenium, lowered in 2023 to 255 µg/day by the European Food Safety Authority,²¹ may still be inadequate to protect against the adverse effects of selenium administration. Conversely, these results are consistent with the WHO recommended dietary intake of 26 and 34 µg/selenium/day among females and males, respectively.⁸²

In general, the results obtained based on achieved blood selenium concentrations at the end of the trial appear to be more reliable than those based on doses of selenium administered in these trials. In fact, the former results took into account compliance with the assigned treatment (selenium supplementation), and individual differences in absorption, excretion, and metabolism of the metalloid. However, when examining differences between achieved blood selenium concentrations and selenium supplementation, we found similar patterns of associations with blood lipid MDs. The only exception was the relationship with LDL cholesterol, for which an opposite effect was observed when considering supplementation doses. Some differences emerged across geographic region, though most of the studies were conducted in Iran, thus limiting the generalizability of our results.

Dose–response analysis based on blood selenium concentrations suggested for 3 lipid end points, though not for HDL cholesterol, a potential increase at very low exposure levels (approximately 50–80 µg/L, ie, 30–50 µg/selenium/day of intake), suggesting there could also be a deleterious effect of selenium deficiency on lipid profile. For HDL cholesterol, however, no such relationship emerged, since the inverse relationship between selenium status and this end point emerged at blood concentrations as low as 50 µg/L.

Observational studies, particularly cross-sectional studies, appear to support the results of our meta-analysis of the experimental data. In fact, positive associations between selenium and metabolic syndrome, higher triglycerides, and LDL cholesterol, as well as with lower HDL cholesterol have previously been reported in the National Health and Nutrition Examination Surveys (NHANES) and other studies.^28,83–90 However, increased serum selenium levels were inversely associated with the average changes in total and LDL cholesterol in a few studies.⁹¹ Similarly, selenium levels demonstrated modestly beneficial effects on blood lipid levels in populations with relatively low selenium status, where low selenium concentrations showed adverse effects, especially on the HDL profile.^92,93

These results are also consistent with the findings from several experimental and nonexperimental human studies indicating adverse metabolic effects of selenium exposure. RCTs have consistently shown that selenium administration is associated with an increased risk of type 2 diabetes mellitus.^21,24,26,94 There is also evidence from observational studies of a detrimental effect of selenium on non-alcoholic fatty acid liver disease,²⁷ on the risk of obesity,²² and on both diastolic and systolic blood pressure levels.^25,31

There is some biological plausibility for an adverse effect of selenium on the lipid profile. High doses of selenium may alter lipid metabolism in laboratory animals by modifying hepatic fatty acid metabolism, protein synthesis, and energy metabolism, and causing increases in body mass.^95–97 In such studies, nutritional levels of selenium also increased triglyceride and total cholesterol levels, with transcriptomic analyses revealing that lipid metabolism was the main pathway altered by dietary selenium treatment. These observations suggest that nutritionally adequate levels of selenium may increase liver lipid contents by activating fatty acid biosynthesis and elongation and unsaturated fatty acid biosynthesis pathways.^95,98 Alteration of redox homeostasis and pro-oxidant features of selenium and selected selenium compounds might represent other pathways involved in such processes.^{12,95,99–101} Finally, too high levels of selenoprotein P, whose excess and adverse effects on glucose metabolism have been involved in the etiology of type 2 diabetes, could also be detrimental to the lipid profile.^29,102

Recent studies indicate that selenium supplementation among individuals with adequate selenium intake could lead to altered energy regulation and cause liver and cardiovascular dysfunction.^95,103 It is feasible that U-shaped associations characterize the relationship between a nutrient such as selenium and the lipid profile, similarly to what has been observed for other outcomes^{25,27,104–108} and in the current meta-analysis. Therefore, selenium deficiency might be linked to inadequate selenoprotein synthesis, potentially leading to disturbances in the redox balance, changes in protein function, and abnormalities in crucial cardiovascular lipid signaling pathways. However, elevated blood selenium concentrations could also be associated with selenoprotein upregulation as a compensatory reaction against the pro-oxidant effects of selenium, thereby having unfavorable effects on certain lipid signaling pathways.^3,8,12,102

RCTs are experimental studies representing the gold standard and can provide the highest quality evidence in scientific research, mainly due to avoidance of confounding via randomization, better exposure assessment, and clarification of temporality. However, most trials conducted so far recruited participants with heterogeneous health status and background selenium concentrations, and administered different amounts of daily selenium supplements. This indicates potential for bias of meta-analyses based on forest plots comparing the highest vs lowest amounts of exposure, given the possibility of U-shaped and J-shaped relationships and the different population characteristics, thus undermining the reliability of both the individual and summary estimates, as shown for other outcomes.^109,110 Therefore, only a dose–response meta-analysis of results, like the one presented in this study, across the entire range of exposure from all trials allows for a comprehensive analysis of the effects of selenium administration on lipid profiles. These aspects should also be taken into account when comparing the results obtained in European populations with those in Asian populations, which could be attributed to underlying differences in selenium status or a specific selenium species. On the other hand, RCTs carried out in Asian populations used a narrower range of selenium supplementation compared with those carried out in European populations, thus limiting their comparability. It is also possible that different confounders, effect mediators, may explain the differences we observed between studies carried out in Asian populations compared with Western populations.

Some limitations of our study must be acknowledged. First, some stratified analyses (eg, by age group) could not be performed, since the trials included did not report MDs according to subgroups, and those that we could carry out yielded imprecise estimates, due to small numbers of studies or study heterogeneity. In addition, none of the studies performed speciation analyses to address the effects on different blood selenium compound concentrations, and few documented the exact chemical forms used for the intervention. Therefore, the extent to which effects differed according to specific selenium species, with reference both to added selenium and to the background status could not be assessed. Sample overlapping (double-counting) for studies combining more than 2 groups is another potential source of bias, but only in forest plots. Finally, many of the included studies were performed among participants with chronic disease and Asian populations, as well as among pregnant females, thus also limiting the generalizability of our results to other populations, though giving on the other hand more reliable evidence for the populations investigated.

CONCLUSIONS

The results of this first dose–response meta-analysis on the association between selenium and lipid profile suggest that selenium levels are associated with adverse effects on lipid profiles, with a nonlinear, generally U-shaped, association, as expected for a substance with both nutritional and toxicological properties. Specifically, selenium doses exceeding 200 µg/day or blood selenium concentrations exceeding 150 µg/L seem to induce adverse effects on the different lipid end points, suggesting that current upper levels of the element may still not be cautious enough, despite their recent updates.^3,21,111 A too-low selenium intake could also be detrimental to lipid metabolism and should be avoided. Overall, these results should be taken into account when considering selenium supplementation for individuals already meeting the nutritional requirements for this element.

Acknowledgments

 

Author Contributions

T.U.: Data curation, Formal analysis, Investigation, Writing—original draft, and Writing—reviewing and editing. G.F.: Data curation, Formal analysis, Writing—original draft, and Writing—reviewing and editing. L.A.W.: Supervision, Writing—original draft, and Writing—reviewing and editing. M.V.: Conceptualization, Supervision, Writing—original draft, and Writing—reviewing and editing. T.F.: Conceptualization, Methodology, Software, Supervision, Validation, Writing—original draft, and Writing—reviewing and editing. All authors have read and approved the final manuscript.

Supplementary Material

Supplementary Material is available at Nutrition Reviews online.

Funding

This study was supported by grant “UNIMORE FAR 2023” funded by the University of Modena and Reggio Emilia. M.V. was supported by the “Fondazione Pietro Manodori” of Reggio Emilia and by funding from the Local Health Authority of Reggio Emilia.

Conflicts of Interest

L.A.W. receives in-kind donations from Kindara.com (fertility apps) and Swiss Precision Diagnostics (home pregnancy tests). She also serves as a consultant for the Gates Foundation and AbbVie, Inc. All of these relationships are for work unrelated to this manuscript. The remaining authors have no relevant interests to declare.

Data Availability

Data described in the manuscript, code book, and analytic code will be made available upon reasonable request, pending application to and approval by the corresponding author.

REFERENCES